Dissecting the Causal Pathway from the Gut Microbiota to PCOS via Circulating Metabolite Mediation: Evidence from Mendelian Randomization, Single-cell and Bulk RNA Sequencing, Network Pharmacology, and In Vitro Tissue Validation

Backgroud Alterations in the gut microbiota have been increasingly linked to the pathogenesis of polycystic ovary syndrome (PCOS), but the causal relationships and underlying molecular mechanisms remain poorly understood. Methods We employed a bidirectional two-sample Mendelian randomization (MR) approach to assess causal associations among the gut microbiota, blood metabolites, and PCOS, utilizing large-scale genome-wide association study (GWAS) summary statistics. Mediation and reverse MR analyses were conducted to identify key metabolic intermediates. The integration of tissue-specific expression quantitative trait locus (eQTL) data facilitated the mapping of microbial genetic variants to ovarian gene regulation. Furthermore, granulosa cells were collected from IVF-ET patients and analyzed via qRT‒PCR to validate the eQTL mapping results. Functional characterization was carried out via the use of bulk and single-cell RNA-sequencing datasets, along with pathway enrichment and immune infiltration analyses. Molecular docking was performed to predict potential drug‒target interactions. Results MR analysis revealed nine gut microbial taxa with suggestive or significant causal associations with PCOS risk, notably a positive effect on Burkholderiales and protective effects on several commensal taxa. The mediation MR identified pseudouridine as a key metabolite mediating the link between Burkholderiales and PCOS. Integration with ovary-specific eQTL data highlighted three candidate genes— ATP13A4 , IMMP2L , and KCTD1 —as potential mediators of microbiota–host interactions. qRT‒PCR analysis of granulosa cells confirmed the high expression of these genes, with ATP13A4 significantly upregulated in PCOS granulosa cells, whereas the other two genes presented significantly lower expression. These findings were supported by strong colocalization signals and validated through bulk and single-cell transcriptomic analyses. Pathway enrichment analyses implicated these genes in inflammatory, metabolic, and hormonal regulation. Immune infiltration and metabolic pathway analyses further elucidated the microenvironmental and functional relevance of these genes in PCOS. Molecular docking has been used to predict drug–gene interactions with potential therapeutic implications. Conclusion Our integrative multiomics MR framework provides robust evidence for a causal gut microbiota–metabolite–gene axis in PCOS pathogenesis. These findings offer new insights into the mechanisms of PCOS and identify novel molecular and pharmacological targets, laying the foundation for the development of precision therapies.